[Gluster-users] software raid performance

baggio liu baggioss at gmail.com
Thu Jun 12 19:01:04 UTC 2008


Hi Matt,
     As you config file, client and server are located at the same machine ,
and they use tcp socket for communication, unix socket may be some
improvement in performance .

:D

Thanks
Baggio


2008/6/13 Daniel Jordan Bambach <dan at lateral.net>:

> Heya Matt,
>
> as per Anands mail to me on the 5th June, much performance can be gained by
> combining the server definitions into the client config and not running
> glusterfsd at all.
>
> I set this up and saw some improvements, but unfortunately there is still
> an outstanding issue about getting the two machines to connect to each other
> properly - I was hoping that Anand might be able to help out on that, but he
> doesn't seem to have been around recently.
>
> D.
>
>
>
> On 12 Jun 2008, at 17:33, Matt Brown wrote:
>
>  On Wednesday 11 June 2008 06:04:16 pm baggio liu wrote:
>>
>>> Hi Matt:
>>>     Can I have the detail comfig of your cluster?? like node number,
>>> config of client and server ??  and so on.
>>>
>>
>> I have two machines, both acting as clients & servers. Below are the fully
>> loaded configs, using server side unify over afr with various performance
>> translators. As you'll see I'm using glusterfs for the /home dir. In users
>> home directories I have php based websites that are being served by apache
>> .
>> The loading of the web pages is very slow compared to using non-gluster
>> file
>> systems on the same machines. I was concerned that perhaps using glusterfs
>> on
>> top of software raid might be the issue. But if that's known to *not*
>> significantly impact performance, then perhaps I should be looking at
>> another
>> way of building this config? Thanks for any insight.
>> Matt
>>
>> # machine 1 server config
>> volume home-web1-ds
>>       type storage/posix
>>       option directory /opt/gluster/home
>> end-volume
>>
>> volume home-web1-ns
>>       type storage/posix
>>       option directory /opt/gluster/home-ns
>> end-volume
>>
>> volume home-web2-ds
>>       type protocol/client
>>       option transport-type tcp/client
>>       option remote-host 10.0.3.22
>>       option remote-subvolume home-web2-ds
>> end-volume
>>
>> volume home-web2-ns
>>       type protocol/client
>>       option transport-type tcp/client
>>       option remote-host 10.0.3.22
>>       option remote-subvolume home-web2-ns
>> end-volume
>>
>> volume home-ns-afr
>>       type cluster/afr
>>       option read-subvolume home-web1-ns
>>       subvolumes home-web1-ns home-web2-ns
>> end-volume
>>
>> volume home-ds-afr
>>       type cluster/afr
>>       option read-subvolume home-web1-ds
>>       subvolumes home-web1-ds home-web2-ds
>> end-volume
>>
>> volume home-unify
>>       type cluster/unify
>>       subvolumes home-ds-afr
>>       option namespace home-ns-afr
>>       option scheduler rr
>> end-volume
>>
>> volume home
>>       type performance/io-threads
>>       option thread-count 2
>>       option cache-size 64MB
>>       subvolumes home-unify
>> end-volume
>>
>> volume server
>>  type protocol/server
>>  option transport-type tcp/server
>>  subvolumes home
>>  option auth.ip.home-web1-ds.allow 10.0.3.*,127.0.0.1
>>  option auth.ip.home-web1-ns.allow 10.0.3.*,127.0.0.1
>>  option auth.ip.home.allow 10.0.3.*,127.0.0.1
>> end-volume
>>
>> # machine 1 client config
>> volume home-mount
>>  type protocol/client
>>  option transport-type tcp/client
>>  option remote-host 127.0.0.1
>>  option remote-subvolume home
>> end-volume
>>
>> volume ioc
>>  type performance/io-cache
>>  subvolumes home-mount
>>  option page-size 1MB
>>  option cache-size 128MB
>>  option force-revalidate-timeout 10
>> end-volume
>>
>> volume writeback
>>  type performance/write-behind
>>  option aggregate-size 1MB
>>  subvolumes ioc
>> end-volume
>>
>> volume readahead
>>  type performance/read-ahead
>>  option page-size 1MB
>>  option page-count 4
>>  option force-atime-update no
>>  subvolumes writeback
>> end-volume
>>
>>
>> # server 2 server config
>> volume home-web2-ds
>>       type storage/posix
>>       option directory /opt/gluster/home
>> end-volume
>>
>> volume home-web2-ns
>>       type storage/posix
>>       option directory /opt/gluster/home-ns
>> end-volume
>>
>> volume home-web1-ds
>>       type protocol/client
>>       option transport-type tcp/client
>>       option remote-host 10.0.3.21
>>       option remote-subvolume home-web1-ds
>> end-volume
>>
>> volume home-web1-ns
>>       type protocol/client
>>       option transport-type tcp/client
>>       option remote-host 10.0.3.21
>>       option remote-subvolume home-web1-ns
>> end-volume
>>
>> volume home-ns-afr
>>       type cluster/afr
>>       option read-subvolume home-web2-ns
>>       subvolumes home-web1-ns home-web2-ns
>> end-volume
>>
>> volume home-ds-afr
>>       type cluster/afr
>>       option read-subvolume home-web2-ds
>>       subvolumes home-web1-ds home-web2-ds
>> end-volume
>>
>> volume home-unify
>>       type cluster/unify
>>       subvolumes home-ds-afr
>>       option namespace home-ns-afr
>>       option scheduler rr
>> end-volume
>>
>> volume home
>>       type performance/io-threads
>>       option thread-count 2
>>       option cache-size 64MB
>>       subvolumes home-unify
>> end-volume
>>
>> volume server
>>  type protocol/server
>>  option transport-type tcp/server
>>  subvolumes home
>>  option auth.ip.home-web2-ds.allow 10.0.3.*,127.0.0.1
>>  option auth.ip.home-web2-ns.allow 10.0.3.*,127.0.0.1
>>  option auth.ip.home.allow 10.0.3.*,127.0.0.1
>> end-volume
>>
>> # sever 2 client config
>> volume home-mount
>>  type protocol/client
>>  option transport-type tcp/client
>>  option remote-host 127.0.0.1
>>  option remote-subvolume home
>> end-volume
>>
>> volume ioc
>>  type performance/io-cache
>>  subvolumes home-mount
>>  option page-size 1MB
>>  option cache-size 128MB
>>  option force-revalidate-timeout 10
>> end-volume
>>
>> volume writeback
>>  type performance/write-behind
>>  option aggregate-size 1MB
>>  subvolumes ioc
>> end-volume
>>
>> volume readahead
>>  type performance/read-ahead
>>  option page-size 1MB
>>  option page-count 4
>>  option force-atime-update no
>>  subvolumes writeback
>> end-volume
>>
>> _______________________________________________
>> Gluster-users mailing list
>> Gluster-users at gluster.org
>> http://zresearch.com/cgi-bin/mailman/listinfo/gluster-users
>>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://supercolony.gluster.org/pipermail/gluster-users/attachments/20080613/d61c4365/attachment.html>


More information about the Gluster-users mailing list